1. 5
  1.  

  2. 1

    This is a problem I have with a lot of the academic papers published by companies, even more than by academics. Google publishes a ton of papers, for example, but many are totally unreproducible without the data used to achieve the results, which they aren’t willing to release. A large number of their machine-translation papers can’t be verified unless you have the Google Translate dataset, for example.